Add support for Templates in OpenShift recipes

[amisevsk]

What does this PR do?

Add support for Templates as OpenShift recipes, with parameters

• If the parameter does not have a default value, it is substituted with
  an empty string
• If the parameter has a default value, it is substituted when launching
  the workspace
• Generated parameter values are not supported (see NoClassDefFoundError when trying to load an OpenShift template with generated parameters fabric8io/kubernetes-client#1340)

This commit does not add dashboard validation (separate PR: #12582), so starting a workspace based off a template requires manually editing an existing workspace config rather than creating a new stack.

The actual change is quite simple: the actual code difference amounts to

clientFactory.create().load(new ByteArrayInputStream(content.getBytes())).get()

instead of

clientFactory.create().lists().load(new ByteArrayInputStream(content.getBytes())).get()

Sample recipe:

kind: Template
apiVersion: v1
metadata:
  name: template-test
objects:
- 
  apiVersion: v1
  kind: Pod
  metadata:
    name: ws
    labels:
      parameter: ${TEST}
  spec:
    containers:
    - 
      image: 'eclipse/che-dev:nightly'
      name: dev
      resources:
        limits:
        memory: 512Mi
parameters:
- name: TEST
  value: test

Above recipe formatted for pasting into existing config:

kind: Template\napiVersion: v1\nmetadata:\n  name: template-test\nobjects:\n- \n  apiVersion: v1\n  kind: Pod\n  metadata:\n    name: ws\n    labels:\n      parameter: ${TEST}\n  spec:\n    containers:\n    - \n      image: 'eclipse/che-dev:nightly'\n      name: dev\n      resources:\n        limits:\n        memory: 512Mi\nparameters:\n- name: TEST\n  value: test

What issues does this PR fix or reference?

#11504

Release Notes

Support OpenShift templates in workspace stack recipes.

Docs PR

Still TODO: docs need to be updated with routes/secrets/templates/etc. notes.

[sleshchenko]

LGTM. Nice to see templates supported by Kubernetes/OpenShfit infrastructure 👍

A small remark to the recipe for testing: indentation is missing before memory field. Resources should be

      resources:
        limits:
          memory: 512Mi

[sleshchenko]

@amisevsk Note that Kubernetes/OpenShfit recipe is also parsed as the list here https://github.com/eclipse/che/blob/master/wsmaster/che-core-api-devfile/src/main/java/org/eclipse/che/api/devfile/server/DevfileEnvironmentFactory.java#L122
Please check if it can be easily handled there and fix/create a separate issue.

[amisevsk]

@sleshchenko I'm not terribly familiar with the devfile codepaths, but it looks like the task would be more complicated there, since it specifically outputs Kubernetes List yaml -- even if we could unmarshal lists/templates, it's not clear how to preserve this when converting back to yaml. I'll create an issue.

[amisevsk]

ci-test

[amisevsk]

Issue here: #12592

[che-bot]

Results of automated E2E tests of Eclipse Che Multiuser on OCP:
Build details
Test report
docker image: eclipseche/che-server:12583
https://github.com/orgs/eclipse/teams/eclipse-che-qa please check this report.

[ibuziuk]

@vkuznyetsov @rhopp could we please get QA approval before merging to master ?

[artaleks9]

Selenium tests execution on Eclipse Che Multiuser on OCP (https://ci.codenvycorp.com/job/che-pullrequests-test-ocp/1511/) doesn't show any regression against this Pull Request.

[dmytro-ndp]

looks like a typo

[amisevsk]

Accidental paste -- thanks!

[dmytro-ndp]

ci-test

[riuvshin]

@dmytro-ndp do you think that comment fix in the unit test requires whole e2e cycle rerun?

[riuvshin]

@amisevsk feel free to merge this right away, do not wait for e2e tests report after comment change...

[che-bot]

Results of automated E2E tests of Eclipse Che Multiuser on OCP:
Build details
Test report
docker image: eclipseche/che-server:12583
https://github.com/orgs/eclipse/teams/eclipse-che-qa please check this report.

[dmytro-ndp]

@riuvshin:

@dmytro-ndp do you think that comment fix in the unit test requires whole e2e cycle rerun?

No, I don't think so.
I had run e2e tests again because there initial commit was rewritten with new one.